Wavelet-based automatic segmentation of the vertebral bodies in digital radiographs

This paper proposes a multi-resolution analysis (MRA) approach to vertebral body segmentation in digital radiographs. A wavelet-based technique carries out the extraction of the most significant features in the radiographs. It generates a multi-resolution edge characterization, discarding noisy artifacts in the meantime. Specific features classification and geometric alignment is then performed, using statistical data to build the. models and curvature analysis to sort the edges. An active contour scheme generates the final contours to be used in clinical studies of the spine.